Re: [PATCH 3/5] drm/xe/xe_ras: Add support to query page offline queue and list

["Tauro, Riana" <riana.tauro@intel.com>]

On 02-09-2026 15:26, Mallesh, Koujalagi wrote:
>
>
> On 25-08-2026 12:06 pm, Riana Tauro wrote:
>> Add support to query page offline list and queue from firmware
>> during module load. The page offline list command retrieves pages that
>> are already offlined by the firmware. The page offline queue command
>> retrieves the pages pending to be offlined by the firmware.
>>
>> Cc: Tejas Upadhyay<tejas.upadhyay@intel.com>
>> Cc: Himal Prasad Ghimiray<himal.prasad.ghimiray@intel.com>
>> Signed-off-by: Riana Tauro<riana.tauro@intel.com>
>> ---
>>   drivers/gpu/drm/xe/xe_ras.c                   | 99 +++++++++++++++++++
>>   drivers/gpu/drm/xe/xe_ras_types.h             | 43 ++++++++
>>   drivers/gpu/drm/xe/xe_sysctrl_mailbox_types.h |  4 +
>>   3 files changed, 146 insertions(+)
>>
>> diff --git a/drivers/gpu/drm/xe/xe_ras.c b/drivers/gpu/drm/xe/xe_ras.c
>> index c643c7137a42..441462a36dbc 100644
>> --- a/drivers/gpu/drm/xe/xe_ras.c
>> +++ b/drivers/gpu/drm/xe/xe_ras.c
>> @@ -328,6 +328,102 @@ static bool ras_counter_is_valid(struct xe_device *xe, struct xe_ras_error_class
>>   	return true;
>>   }
>>   
>> +static void get_queued_pages(struct xe_device *xe)
>> +{
>> +	struct xe_sysctrl_mailbox_command command = {0};
>> +	struct xe_ras_page_offline_queue response = {0};
>> +	u32 count = 0;
>> +	size_t rlen;
>> +	int ret, i;
>> +
>> +	/* Supported only on platforms with system controller */
>> +	if (!xe->info.has_sysctrl)
>> +		return;
>> +
>> +	xe_sysctrl_create_command(&command, XE_SYSCTRL_GROUP_GFSP,
>> +				  XE_SYSCTRL_CMD_GET_OFFLINE_QUEUE, NULL, 0, &response,
>> +				  sizeof(response));
>> +
>> +	do {
>> +		memset(&response, 0, sizeof(response));
>> +
>> +		ret = xe_sysctrl_send_command(&xe->sc, &command, &rlen);
>> +		if (ret) {
>> +			xe_log_err_fatal(xe, SYSCTRL, ret, "failed to get page offline queue\n");
>> +			return;
>> +		}
>> +		if (rlen != sizeof(response)) {
>> +			xe_log_err(xe, SYSCTRL, -EINVAL,
> may be use errno -EPROTO?

This is consistent with the rest of the code in the file

>> +				   "unexpected page offline queue response length %zu (expected %zu)\n",
>> +				   rlen, sizeof(response));
>> +			return;
>> +		}
>> +
>> +		for (i = 0; i < response.pages_returned && i < XE_RAS_NUM_PAGES; i++)
>> +			handle_page_offline(xe, response.page_addresses[i], true);
> Silently dropping errors from handle_page_offline (). Should handle 
> errors.


The only errors are critical bo's which won't be possible at boot and 
system controller failure.
System controller failure could be due to multiple reasons and it was 
decided that driver need not wedge for this on boot.
Any missed memory error will be reported again as AER

Logging is already present.

>
>> +
>> +		count += response.pages_returned;
>> +		if (!response.pages_returned)
>> +			break;
>> +
> To avoid infinite loop due to bad firmware use flood limit right?

We do have the below

>> +		if (count > response.total_pages) {
>> +			xe_log_err(xe, SYSCTRL, -EINVAL,
>> +				   "Pages returned from queue exceed total pages %u, returned %u\n",
>> +				   response.total_pages, count);
>> +			return;
>> +		}
>> +	} while (response.additional_data);
>> +}
>> +
>> +static void get_offlined_list(struct xe_device *xe)
>> +{
>> +	struct xe_sysctrl_mailbox_command command = {0};
>> +	struct xe_ras_offline_list_response response = {0};
>> +	struct xe_ras_offline_list_request request = {0};
>> +	u32 count = 0;
>> +	size_t rlen;
>> +	int ret, i;
>> +
>> +	/* Supported only on platforms with system controller */
>> +	if (!xe->info.has_sysctrl)
>> +		return;
>> +
>> +	xe_sysctrl_create_command(&command, XE_SYSCTRL_GROUP_GFSP, XE_SYSCTRL_CMD_GET_OFFLINE_LIST,
>> +				  &request, sizeof(request), &response, sizeof(response));
>> +
>> +	do {
>> +		memset(&response, 0, sizeof(response));
>> +		request.index = count;
>> +
>> +		ret = xe_sysctrl_send_command(&xe->sc, &command, &rlen);
>> +		if (ret) {
>> +			xe_log_err_fatal(xe, SYSCTRL, ret, "failed to get page offline list\n");
>> +			return;
>> +		}
>> +
>> +		if (rlen != sizeof(response)) {
>> +			xe_log_err(xe, SYSCTRL, -EINVAL,
> may be use errno -EPROTO?

as above

>> +				   "unexpected page offline list response length %zu (expected %zu)\n",
>> +				   rlen, sizeof(response));
>> +			return;
>> +		}
>> +
>> +		for (i = 0; i < response.pages_returned && i < XE_RAS_NUM_PAGES; i++)
>> +			handle_page_offline(xe, response.page_addresses[i], false);
>> +
> Silently dropping errors, need to handle?
>

Same as above

Thanks
Riana


>> +		count += response.pages_returned;
>> +		if (!response.pages_returned)
>> +			break;
>> +
> To avoid infinite loop due to bad firmware use flood limit right?
>> +		if (count > response.total_pages) {
>> +			xe_log_err(xe, SYSCTRL, -EINVAL,
>> +				   "Pages returned from list exceed total pages %u, returned %u\n",
>> +				   response.total_pages, count);
>> +			return;
>> +		}
>> +	} while (response.additional_data);
>> +}
>> +
>>   static struct pci_dev *find_usp_dev(struct pci_dev *pdev)
>>   {
>>   	struct pci_dev *vsp;
>> @@ -923,6 +1019,9 @@ void xe_ras_init(struct xe_device *xe)
>>   	if (IS_ENABLED(CONFIG_PCIEAER))
>>   		ras_usp_aer_init(xe);
>>   
>> +	get_queued_pages(xe);
> Better to handle errors rigtht?
>> +	get_offlined_list(xe);
> ditto
>> +
>>   	ret = devm_device_add_group(xe->drm.dev, &gpu_health_group);
>>   	if (ret)
>>   		xe_err(xe, "Failed to create GPU health sysfs, err=%d\n", ret);
>> diff --git a/drivers/gpu/drm/xe/xe_ras_types.h b/drivers/gpu/drm/xe/xe_ras_types.h
>> index 2fac968879b6..cddcfa656d9f 100644
>> --- a/drivers/gpu/drm/xe/xe_ras_types.h
>> +++ b/drivers/gpu/drm/xe/xe_ras_types.h
>> @@ -10,6 +10,7 @@
>>   
>>   #define XE_RAS_NUM_COUNTERS			16
>>   #define XE_RAS_NUM_ERROR_ARR			3
>> +#define XE_RAS_NUM_PAGES			25
>>   /* Error bits in IEH global error status register */
>>   #define XE_RAS_SOC_IEH_PUNIT			BIT(1)
>>   /* Device memory error categories */
>> @@ -280,6 +281,48 @@ struct xe_ras_page_offline_response {
>>   	u32 reserved;
>>   } __packed;
>>   
>> +/**
>> + * struct xe_ras_offline_list_request - Request for get offline list command
>> + */
>> +struct xe_ras_offline_list_request {
>> +	/** @index: Zero-based index into the offline page list */
>> +	u32 index;
>> +} __packed;
>> +
>> +/**
>> + * struct xe_ras_offline_list_response - Response from get offline list command
>> + */
>> +struct xe_ras_offline_list_response {
>> +	/** @max_entries: Total no of pages that can be stored in flash */
>> +	u32 max_entries;
>> +	/** @total_pages: Total number of permanently offlined pages */
>> +	u32 total_pages;
>> +	/** @pages_returned: Number of pages returned in this response */
>> +	u32 pages_returned;
>> +	/** @page_addresses: Array of permanently offlined page addresses (4KB aligned) */
>> +	u64 page_addresses[XE_RAS_NUM_PAGES];
>> +	/** @additional_data: Indicates if more data is available */
>> +	u8 additional_data;
>> +	/** @reserved: Reserved for future use */
>> +	u8 reserved[3];
>> +} __packed;
>> +
>> +/**
>> + * struct xe_ras_page_offline_queue - Response from get offline queue command
>> + */
>> +struct xe_ras_page_offline_queue {
>> +	/** @total_pages: Total number of queued pages */
>> +	u32 total_pages;
>> +	/** @pages_returned: Number of pages returned in this response */
>> +	u32 pages_returned;
>> +	/** @page_addresses: Array of page addresses (4KB aligned) */
>> +	u64 page_addresses[XE_RAS_NUM_PAGES];
>> +	/** @additional_data: Indicates if more data is available */
>> +	u8 additional_data;
>> +	/** @reserved: Reserved for future use */
>> +	u8 reserved[3];
>> +} __packed;
>> +
>>   /**
>>    * struct xe_ras_get_health_request - Request structure for obtaining gpu health
>>    */
>> diff --git a/drivers/gpu/drm/xe/xe_sysctrl_mailbox_types.h b/drivers/gpu/drm/xe/xe_sysctrl_mailbox_types.h
>> index 3363f48da2b7..194ad3ac3da2 100644
>> --- a/drivers/gpu/drm/xe/xe_sysctrl_mailbox_types.h
>> +++ b/drivers/gpu/drm/xe/xe_sysctrl_mailbox_types.h
>> @@ -27,6 +27,8 @@ enum xe_sysctrl_group {
>>    * @XE_SYSCTRL_CMD_CLEAR_COUNTER: Clear error counter value
>>    * @XE_SYSCTRL_CMD_GET_PENDING_EVENT: Retrieve pending event
>>    * @XE_SYSCTRL_CMD_PAGE_OFFLINE: Instruct firmware to offline/decline a page
>> + * @XE_SYSCTRL_CMD_GET_OFFLINE_LIST: Retrieve list of all offlined pages from flash
>> + * @XE_SYSCTRL_CMD_GET_OFFLINE_QUEUE: Retrieve list of offlined queued pages from firmware
>>    * @XE_SYSCTRL_CMD_GET_HEALTH: Retrieve gpu health
>>    * @XE_SYSCTRL_CMD_SET_HEALTH: Set gpu health
>>    */
>> @@ -36,6 +38,8 @@ enum xe_sysctrl_gfsp_cmd {
>>   	XE_SYSCTRL_CMD_CLEAR_COUNTER		= 0x04,
>>   	XE_SYSCTRL_CMD_GET_PENDING_EVENT	= 0x07,
>>   	XE_SYSCTRL_CMD_PAGE_OFFLINE             = 0x08,
>> +	XE_SYSCTRL_CMD_GET_OFFLINE_LIST		= 0x09,
>> +	XE_SYSCTRL_CMD_GET_OFFLINE_QUEUE	= 0x0A,
>>   	XE_SYSCTRL_CMD_GET_HEALTH		= 0x0B,
>>   	XE_SYSCTRL_CMD_SET_HEALTH		= 0x0C,
>>   };